Bayesian Model of Polygonal Chain for Landmark Point Detection

• Activity Number: 360 - Contributed Poster Presentations: Section on Bayesian Statistical Science
• Type: Contributed
• Date/Time: Wednesday, August 5, 2020 : 10:00 AM to 2:00 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Cong Zhang* and Min Chen and Kelli Palmer and Michael Zhang and Qiwei Li
• Companies: University of Texas at Dallas and University of Texas at Dallas and University of Texas at Dallas and University of Texas at Dallas and The University of Texas at Dallas
• Keywords: Bayesian; tumor; shape; landmark; polygonal chain; pathology image

Abstract

Medical image has become a pivotal routine tool to help the diagnosis, prognosis prediction and treatment of tumor with the rapid advance of medical imaging technology. Tumor speculation indicates invasive spread and advanced stages while less aggressive tumors often have well-defined margins. Recent studies based on tumor shape features has exhibited great potential in tumor diagnosis and outcome prediction while most of those features are indicative such as tumor circularity or depend on subjective pre-defined landmarks, lacking underlying statistical model for characterization of tumor shape and uncertainty quantification. Here, we consider the tumor margins as polygonal chains and propose a Bayesian framework which identifies landmark points to characterize tumor shape and provide descriptive statistic for following analysis. Markov chain Monte Carlo sampling approach and Metropolis–Hastings algorithm were used for model fitting. The proposed model was applied to the pathology images of 268 lung cancer patients from the National Lung Screening trial and obtained landmark points approximates polygonal chains of tumor regions. Those landmark points can be used to generate feature